其他
故障解析丨导入字符串NULL导致主从报错
1.背景概述
目前需要搭建一个从库，由于单表数据量较大，时间比较有限，考虑到导入导出的时间，并且GreatSQL支持并行load data的功能，能够加速数据的导入，因此决定使用 select into outfile 和 load data 的方式进行数据的迁移；
在数据导入完成后进行数据同步，从库发生报错 1032 找不到记录，但是登录从库中可以查询到此条记录，这里就很奇怪；
最后通过解析relaylog，根据relaylog中的update记录，以每个字段为查询条件进行查询，发现是由于NULL值列导致的，主库这列的值是 NULL，从库在导入后导成了字符串"NULL"，因此导致回放update操作时匹配不到数据而报错1032.
2.问题复现
本次测试基于 GreatSQL 8.0.32
2.1 初始化2个单机实例
略
2.2 主节点创建测试表
greatsql> create database test;
greatsql> use test;
greatsql> create table t1 (id int,
name varchar(30),
age int,
addr varchar(30),
school varchar(30),
unique key (id)) engine=innodb;
greatsql> insert into t1 values
(1,'小红',10,'北京','一中'),
(2,'小绿',11,'北京','一中'),
(3,'小黄',12,'北京',NULL),
(4,'小蓝',13,'北京',NULL),
(5,'小黑',14,'北京',NULL);
2.3 查看数据
greatsql> select * from t1;
+----+--------+------+--------+--------+
| id | name | age | addr | school |
+----+--------+------+--------+--------+
| 1 | 小红 | 10 | 北京 | 一中 |
| 2 | 小绿 | 11 | 北京 | 一中 |
| 3 | 小黄 | 12 | 北京 | NULL |
| 4 | 小蓝 | 13 | 北京 | NULL |
| 5 | 小黑 | 14 | 北京 | NULL |
+----+--------+------+--------+--------+
5 rows in set (0.00 sec)
2.4 主节点导出数据
greatsql> select * from test.t1 into outfile '/greatsql/t1.csv' FIELDS TERMINATED BY '|+|' ESCAPED BY '' LINES TERMINATED BY '/*rowsxxx*/';
2.5 查看导出的数据
$ cat t1.csv
1|+|小红|+|10|+|北京|+|一中/*rowsxxx*/2|+|小绿|+|11|+|北京|+|一中/*rowsxxx*/3|+|小黄|+|12|+|北京|+|NULL/*rowsxxx*/4|+|小蓝|+|13|+|北京|+|NULL/*rowsxxx*/5|+|小黑|+|14|+|北京|+|NULL/*rowsxxx
可以看到导出的数据中包含 NULL
2.6 从库创建表并导入数据
greatsql> create database test;
use test;
create table t1 (id int,
name varchar(30),
age int,
addr varchar(30),
school varchar(30),
unique key (id)) engine=innodb;
导入数据
greatsql> load data infile '/greatsql/t1.csv' into table test.t1 fields terminated by '|+|' ESCAPED BY '' lines terminated by '/*rowsxxx*/';
2.7 从库查询数据
greatsql> select * from test.t1;
+----+--------+------+--------+--------+
| id | name | age | addr | school |
+----+--------+------+--------+--------+
| 1 | 小红 | 10 | 北京 | 一中 |
| 2 | 小绿 | 11 | 北京 | 一中 |
| 3 | 小黄 | 12 | 北京 | NULL |
| 4 | 小蓝 | 13 | 北京 | NULL |
| 5 | 小黑 | 14 | 北京 | NULL |
+----+--------+------+--------+--------+
5 rows in set (0.00 sec)
2.8 从库建立复制
greatsql> reset master;
Query OK, 0 rows affected (0.04 sec)
greatsql> set global gtid_purged='b94e6517-68dd-11ee-b43b-00163ecb92e3:1-5755';
Query OK, 0 rows affected (0.00 sec)
greatsql> show master status;
+---------------+----------+--------------+------------------+---------------------------------------------+
| File | Position | Binlog_Do_DB | Binlog_Ignore_DB | Executed_Gtid_Set |
+---------------+----------+--------------+------------------+---------------------------------------------+
| binlog.000001 | 153 | | | b94e6517-68dd-11ee-b43b-00163ecb92e3:1-5755 |
+---------------+----------+--------------+------------------+---------------------------------------------+
1 row in set (0.00 sec)
greatsql> change master to master_user='root',master_password='greatdb',master_host='192.168.137.162',master_port=6001,master_auto_position=1;
Query OK, 0 rows affected, 7 warnings (0.02 sec)
greatsql> start slave;
Query OK, 0 rows affected, 1 warning (0.04 sec)
greatsql> show slave status\G
*************************** 1. row ***************************
Slave_IO_State: Waiting for source to send event
Master_Host: 192.168.137.162
Master_User: root
Master_Port: 6001
Connect_Retry: 60
Master_Log_File: binlog.000002
Read_Master_Log_Pos: 1861574
Relay_Log_File: relaylog.000002
Relay_Log_Pos: 395
Relay_Master_Log_File: binlog.000002
Slave_IO_Running: Yes
Slave_SQL_Running: Yes
Replicate_Do_DB:
......
2.9 主库更新数据
greatsql> update test.t1 set name='小小黑' where id=5;
2.10 从库查看复制状态
greatsql> show slave status\G
*************************** 1. row ***************************
Slave_IO_State: Waiting for source to send event
Master_Host: 172.17.137.162
Master_User: root
Master_Port: 6001
Connect_Retry: 60
Master_Log_File: binlog.000002
Read_Master_Log_Pos: 1863564
Relay_Log_File: relaylog.000002
Relay_Log_Pos: 395
Relay_Master_Log_File: binlog.000002
Slave_IO_Running: Yes
Slave_SQL_Running: No
Replicate_Do_DB:
Replicate_Ignore_DB:
Replicate_Do_Table:
Replicate_Ignore_Table:
Replicate_Wild_Do_Table:
Replicate_Wild_Ignore_Table:
Last_Errno: 1032
Last_Error: Coordinator stopped because there were error(s) in the worker(s). The most recent failure being: Worker 1 failed executing transaction 'b94e6517-68dd-11ee-b43b-00163ecb92e3:5756' at master log binlog.000002, end_log_pos 1863537. See error log and/or performance_schema.replication_applier_status_by_worker table for more details about this failure or others, if any.
Skip_Counter: 0
......
greatsql> select * from performance_schema.replication_applier_status_by_worker limit 1\G
*************************** 1. row ***************************
CHANNEL_NAME:
WORKER_ID: 1
THREAD_ID: NULL
SERVICE_STATE: OFF
LAST_ERROR_NUMBER: 1032
LAST_ERROR_MESSAGE: Worker 1 failed executing transaction 'b94e6517-68dd-11ee-b43b-00163ecb92e3:5756' at master log binlog.000002, end_log_pos 1863537; Could not execute Update_rows event on table test.t1; Can't find record in 't1', Error_code: 1032; Can't find record in 't1', Error_code: 1032; handler error HA_ERR_KEY_NOT_FOUND; the event's master log FIRST, end_log_pos 1863537
LAST_ERROR_TIMESTAMP: 2023-10-17 10:02:46.396166
LAST_APPLIED_TRANSACTION:
LAST_APPLIED_TRANSACTION_ORIGINAL_COMMIT_TIMESTAMP: 0000-00-00 00:00:00.000000
LAST_APPLIED_TRANSACTION_IMMEDIATE_COMMIT_TIMESTAMP: 0000-00-00 00:00:00.000000
LAST_APPLIED_TRANSACTION_START_APPLY_TIMESTAMP: 0000-00-00 00:00:00.000000
LAST_APPLIED_TRANSACTION_END_APPLY_TIMESTAMP: 0000-00-00 00:00:00.000000
APPLYING_TRANSACTION: b94e6517-68dd-11ee-b43b-00163ecb92e3:5756
APPLYING_TRANSACTION_ORIGINAL_COMMIT_TIMESTAMP: 2023-10-17 10:02:46.392331
APPLYING_TRANSACTION_IMMEDIATE_COMMIT_TIMESTAMP: 2023-10-17 10:02:46.392331
APPLYING_TRANSACTION_START_APPLY_TIMESTAMP: 2023-10-17 10:02:46.393814
LAST_APPLIED_TRANSACTION_RETRIES_COUNT: 0
LAST_APPLIED_TRANSACTION_LAST_TRANSIENT_ERROR_NUMBER: 0
LAST_APPLIED_TRANSACTION_LAST_TRANSIENT_ERROR_MESSAGE:
LAST_APPLIED_TRANSACTION_LAST_TRANSIENT_ERROR_TIMESTAMP: 0000-00-00 00:00:00.000000
APPLYING_TRANSACTION_RETRIES_COUNT: 0
APPLYING_TRANSACTION_LAST_TRANSIENT_ERROR_NUMBER: 0
APPLYING_TRANSACTION_LAST_TRANSIENT_ERROR_MESSAGE:
APPLYING_TRANSACTION_LAST_TRANSIENT_ERROR_TIMESTAMP: 0000-00-00 00:00:00.000000
1 row in set (0.00 sec)
可以看到从库发生1032报错，找不到记录
2.11 解析从库relay log
#231017 10:02:46 server id 1 end_log_pos 1863456 Table_map: `test`.`t1` mapped to number 180
\# has_generated_invisible_primary_key=0
\# at 673
\#231017 10:02:46 server id 1 end_log_pos 1863537 Update_rows: table id 180 flags: STMT_END_F
\### UPDATE `test`.`t1`
\### WHERE
\### @1=5 /* INT meta=0 nullable=1 is_null=0 */
\### @2='小黑' /* VARSTRING(120) meta=120 nullable=1 is_null=0 */
\### @3=14 /* INT meta=0 nullable=1 is_null=0 */
\### @4='北京' /* VARSTRING(120) meta=120 nullable=1 is_null=0 */
\### @5=NULL /* VARSTRING(120) meta=120 nullable=1 is_null=1 */
\### SET
\### @1=5 /* INT meta=0 nullable=1 is_null=0 */
\### @2='小小黑' /* VARSTRING(120) meta=120 nullable=1 is_null=0 */
\### @3=14 /* INT meta=0 nullable=1 is_null=0 */
\### @4='北京' /* VARSTRING(120) meta=120 nullable=1 is_null=0 */
\### @5=NULL /* VARSTRING(120) meta=120 nullable=1 is_null=1 */
\# at 754
\#231017 10:02:46 server id 1 end_log_pos 1863564 Xid = 5940
COMMIT/*!*/;
SET @@SESSION.GTID_NEXT= 'AUTOMATIC' /* added by greatsqlbinlog */ /*!*/;
DELIMITER ;
\# End of log file
可以看到update更新前的数据与从库的数据一致，那么为什么还会报错 找不到记录呢？
2.12 根据relay log中的内容去从库查询数据
greatsql> select * from test.t1 where id=5;
+------+--------+------+--------+--------+
| id | name | age | addr | school |
+------+--------+------+--------+--------+
| 5 | 小黑 | 14 | 北京 | NULL |
+------+--------+------+--------+--------+
1 row in set (0.01 sec)
greatsql> select * from test.t1 where name='小黑';
+------+--------+------+--------+--------+
| id | name | age | addr | school |
+------+--------+------+--------+--------+
| 5 | 小黑 | 14 | 北京 | NULL |
+------+--------+------+--------+--------+
1 row in set (0.01 sec)
greatsql> select * from test.t1 where age=14;
+------+--------+------+--------+--------+
| id | name | age | addr | school |
+------+--------+------+--------+--------+
| 5 | 小黑 | 14 | 北京 | NULL |
+------+--------+------+--------+--------+
1 row in set (0.00 sec)
greatsql> select * from test.t1 where addr='北京';
+------+--------+------+--------+--------+
| id | name | age | addr | school |
+------+--------+------+--------+--------+
| 1 | 小红 | 10 | 北京 | 一中 |
| 2 | 小绿 | 11 | 北京 | 一中 |
| 3 | 小黄 | 12 | 北京 | NULL |
| 4 | 小蓝 | 13 | 北京 | NULL |
| 5 | 小黑 | 14 | 北京 | NULL |
+------+--------+------+--------+--------+
5 rows in set (0.00 sec)
greatsql> select * from test.t1 where school is null;
Empty set (0.01 sec)
greatsql> select * from test.t1 where school='null';
+------+--------+------+--------+--------+
| id | name | age | addr | school |
+------+--------+------+--------+--------+
| 3 | 小黄 | 12 | 北京 | NULL |
| 4 | 小蓝 | 13 | 北京 | NULL |
| 5 | 小黑 | 14 | 北京 | NULL |
+------+--------+------+--------+--------+
3 rows in set (0.00 sec)
可以看到，根据null值作为查询条件时，匹配不到数据； 根据字符串"null" 进行匹配是可以匹配到数据
2.13 去主库进行查询
greatsql> select * from test.t1 where school is null;
+------+-----------+------+--------+--------+
| id | name | age | addr | school |
+------+-----------+------+--------+--------+
| 3 | 小黄 | 12 | 北京 | NULL |
| 4 | 小蓝 | 13 | 北京 | NULL |
| 5 | 小小黑 | 14 | 北京 | NULL |
+------+-----------+------+--------+--------+
3 rows in set (0.00 sec)
greatsql> select * from test.t1 where school='null';
Empty set (0.00 sec)
在主库查询的结果与从库相反
可以得出结论，由于从库导入的数据将NULL值列的数据导入成了字符串 NULL，因此导致主从数据出现了不一致。
2.14 修复从库
greatsql> stop slave;
Query OK, 0 rows affected, 1 warning (0.00 sec)
greatsql> update test.t1 set school=NULL where school='null';
Query OK, 3 rows affected (0.01 sec)
Rows matched: 3 Changed: 3 Warnings: 0
greatsql> start slave;
Query OK, 0 rows affected, 1 warning (0.04 sec)
greatsql> show slave status\G
*************************** 1. row ***************************
Slave_IO_State: Waiting for source to send event
Master_Host: 172.17.137.162
Master_User: root
Master_Port: 6001
Connect_Retry: 60
Master_Log_File: binlog.000002
Read_Master_Log_Pos: 1863564
Relay_Log_File: relaylog.000003
Relay_Log_Pos: 435
Relay_Master_Log_File: binlog.000002
Slave_IO_Running: Yes
Slave_SQL_Running: Yes
Replicate_Do_DB:
......
可以看到主从状态已经恢复正常
3.总结
1.如果FIELDS ESCAPED BY字符为空字符，则没有字符被转义，并且NULL被作为NULL输出，而不是\N；这也是导致此次主从报错的原因。
2.如果这张表使用的是主键而不是唯一索引，即使某些列被导入为字符串NULL，也不会报错。
3.如果这张表没有索引或有普通索引，则会报错。
Enjoy GreatSQL :)
《深入浅出MGR》视频课程
戳此小程序即可直达B站
文章推荐
GreatSQL社区与Amazon、Facebook、Tencent共同被MySQL致谢
GreatSQL登陆Arch Linux：成功的数据库安装之旅
<关于 GreatSQL>
GreatSQL 是适用于金融级应用的国内自主开源数据库，具备高性能、高可靠、高易用性、高安全等多个核心特性，可以作为 MySQL 或 Percona Server 的可选替换，用于线上生产环境，且完全免费并兼容 MySQL 或 Percona Server。
（对文章有疑问或者有独到见解都可以去社区官网提出或分享哦~）
微信&QQ群：可扫码添加GreatSQL社区助手微信好友，发送验证信息“加群”加入GreatSQL/MGR交流微信群，亦可直接扫码加入GreatSQL/MGR交流QQ群。
微信
想看更多技术好文，点个“在看”吧！